ResearchOct 28, 2026, 00:00 UTC

ChronoReplay trains spatial VLMs with chronological replay of heterogeneous supervision

The paper introduces ChronoReplay, which replays spatial supervision as ordered episodes and reports a 13.41-point gain on OST-Bench alongside gains on OVO-S-Bench and UCS-Bench. Adaptive frame selection adds an inference-time gain at lower visual cost.

ModelChronoReplay
Topicresearch
Topicbenchmark

1outlet ran itRead at Hugging Face Papers

Status: Research
Published: Oct 28, 2026, 00:00 UTC That day
Outlets: 1
Written by: Claude How

Every outlet that ran it 1 article

Oct 28, 2026, 00:00 UTC
Hugging Face PapersResearchLead
ChronoReplay: One Answer Interface Is Not Enough for Streaming Spatial Intelligence

All news